Pharmacokinetics-pharmacodynamics population model building with NONMEM on amphetamine’s effect affecting release of dopamine
-
Background Characterizing the relationship between the pharmacokinetics (PK, concentration vs. time) and pharmacodynamics (PD, effect vs. time) is an important tool being used in both pharmaceutical industry and medical institutes. Amphetamine is a central nervous system stimulant, which will causes emotional and cognitive boost and its main elimination route is via kidney, with 40% unchanged drug. As known, amphetamine will cause a release of dopamine in the central nervous system.
-
Aim The objective for this study is to build a population model explaining the relation between amphetamine concentrations and dopamine levels.
-
Dataset used in modeling The dataset, collected from animal experiments, consists of the data of 20 different individuals, with the amphetamine concentration vs. time and dopamine concentration vs. time. Measurements of dopamine levels had been taken in the nucleus accumbens using micro-dialysis, when amphetamine concentrations were measured in plasma. Each individual had their dopamine baseline prior to a 5 mg intraperitoneal amphetamine administration. After administration, both dopamine and amphetamine levels were measured every 15 minutes for 180 minutes after administration.
-
Method
4.1 Tools used in modeling During model building, NONMEM 7.3 was used for modeling and estimation. Meanwhile, PsN and Xpose 4.6.0 in R environment were used to assist operation. After models were finished, Berkeley Madonna v. 8 was used as a tool of plotting prediction.
4.2 Model building and evaluation criteria The models consists of three components. The structural model included the basic properties, e.g. absorption rate, and number of compartments, the statistical model included between-subject, between-occasion, and residual variabilities, and the covariate model included the relationships between parameters. Each model in NONMEM model included a pharmacokinetic code with initial parameter estimates and bounds (THETA values), with between subject variability (ETA values), as well as differential equations to specify amount and rate of transfer between compartments. Furthermore, an error code was included to specify the model for residual variability which specifies how the variability of observed data varies from the individual model predictions. Covariate data (OMEGA values) and variance of the residual variability (SIGMA values) were also included in the model. Models were evaluated by five criteria: justification from literature and dataset statistics, relative change in objective function with statistics significance, uncertainty of parameter estimates, goodness-of-fit plots in R, and simulation based diagnostics.
4.3 PK/PD model For PK model, three parameters were taken into account: clearance (CL) volume of distribution (V) and absorption rate constant (Ka). Furthermore, the drug is considered following first order absorption. At first stage, both one compartment and two compartments model are tested. After deciding model type, between subject variability (BSV) and residual unexplained variability (RUV) were added to each parameters, with consideration of various relation with parameters (exponential, proportional and additive). After comparison and adjustment, the final PK model was decided as first order absorption, one compartment model with exponential BSV on CL and V, proportional RUV. With the estimation of relevant parameters from PK model, PD model was developed. To seek for best fitting modeltype to dataset, five models were tested: direct effect linear, power linear, Emax, sigmoid Emax and indirect effect model. In the following stage, BSV were added. Above all, the model performed best, which is an indirect effect model combined with linear relation of drug concentration, was chosen.
- Results After all the attempts made, a final decision is made: PK model is a first order absorption, one compartment model with exponential BSV on CL and V, proportional RUV. With the data obtained from PK model estimation, the PD model is developed as an indirect effect with linear concentration change with exponential BSV for all parameters and proportional RUV.